Shaping Urban Resilience

In order to shape urban resilience, it is necessary to understand disaster risks to get better disaster response. Twitter allows people to collect abundant real-time or historical social media data via its API, which gradually make it a repository for disaster-related information collection. This study has two research objectives. The first is to evaluate whether Twitter data can reflect the emergency and vulnerability and thus aid in disaster response when Hurricane Harvey struck Houston in 2017. The second is to evaluate whether Twitter data can be used to perform damage assessment after Hurricane Harvey. Three new conceptions are defined to perform evaluation: tweet awareness (TAw), tweet activity (TAc) and tweet focus (TFo). By comparing with other variables such as normalized average proximity (NAP), social vulnerability index (SVI) through spatiotemporal analysis, main conclusions are drawn as the following. First, the temporal distribution of tweets is periodic: the tweets at night are much more than that in the daytime and there exists the “outbreak” time of the tweets. Second, when the hurricane is getting closer to the land, the TAw is increasing and vice versa, which reflect the emergency situation of hurricane temporally. Third, there is no statistically significant relationship between TAw and NAP based on county-level data. Fourth, the relationship between TFo and SVI is not statistically significant and thus, the twitter data could not reflect the social vulnerability. Next, there is no significant relationship between them spatially and it is not feasible to perform rapid assessment of damage loss. Last but not least, the highest clustered point (the points share the same coordinate) of hurricane-related tweets is located in the University of Houston Downtown, which indicates that main active users of Twitter might be college school students.